{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,11]],"date-time":"2026-05-11T22:47:32Z","timestamp":1778539652742,"version":"3.51.4"},"reference-count":55,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2025,6,19]],"date-time":"2025-06-19T00:00:00Z","timestamp":1750291200000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Artif. Intell."],"abstract":"<jats:p>DeepSeek has introduced its recent model DeepSeek-R1, showing divergence from OpenAI\u2019s ChatGPT, suggesting an open-source alternative to users. This paper analyzes the architecture of DeepSeek-R1, mainly adopting rule-based reinforcement learning (RL) without preliminary supervised fine-tuning (SFT), which has shown better efficiency. By integrating multi-stage training along with cold-start data usage before RL, the model can achieve meaningful performance in reasoning tasks along with reward modeling optimizing training process. DeepSeek shows its strength in technical, reasoning tasks, able to show its decision-making process through open source whereas ChatGPT shows its strength on general tasks and areas requiring creativeness. Despite the groundbreaking developments of both models, there is room for improvement in AI landscape and matters to be handled such as quality of data, black box problems, privacy management, and job displacement. This paper suggests the future of AI, expecting better performance in multi-modal tasks, enhancing its effectiveness in handling larger data sets, enabling users with improved AI landscapes and utility.<\/jats:p>","DOI":"10.3389\/frai.2025.1576992","type":"journal-article","created":{"date-parts":[[2025,6,19]],"date-time":"2025-06-19T05:30:05Z","timestamp":1750311005000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":32,"title":["DeepSeek vs. ChatGPT: prospects and challenges"],"prefix":"10.3389","volume":"8","author":[{"given":"Inhye","family":"Jin","sequence":"first","affiliation":[]},{"given":"Jonathan A.","family":"Tangsrivimol","sequence":"additional","affiliation":[]},{"given":"Erfan","family":"Darzi","sequence":"additional","affiliation":[]},{"given":"Hafeez Ul","family":"Hassan Virk","sequence":"additional","affiliation":[]},{"given":"Zhen","family":"Wang","sequence":"additional","affiliation":[]},{"given":"Jan","family":"Egger","sequence":"additional","affiliation":[]},{"given":"Sean","family":"Hacking","sequence":"additional","affiliation":[]},{"given":"Benjamin S.","family":"Glicksberg","sequence":"additional","affiliation":[]},{"given":"Markus","family":"Strauss","sequence":"additional","affiliation":[]},{"given":"Chayakrit","family":"Krittanawong","sequence":"additional","affiliation":[]}],"member":"1965","published-online":{"date-parts":[[2025,6,19]]},"reference":[{"key":"ref1","article-title":"Gpt-4 technical report","volume-title":"arXiv","author":"Achiam","year":"2023"},{"key":"ref2","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/2023.emnlp-main.298","article-title":"GQA: training generalized multi-query transformer models from multi-head checkpoints","volume-title":"arXiv","author":"Ainslie","year":"2023"},{"key":"ref3","doi-asserted-by":"publisher","first-page":"1","DOI":"10.37965\/jait.2025.0740","article-title":"Deepseek vs. ChatGPT: a comparative evaluation of AI tools in composition, business writing, and communication tasks","volume":"4","author":"AlAfnan","year":"2025","journal-title":"J. Artif. Intell. Technol."},{"key":"ref4","year":"2024"},{"key":"ref5","article-title":"Deepseek llm: scaling open-source language models with longtermism","volume-title":"arXiv","author":"Bi","year":"2024"},{"key":"ref6","article-title":"Language models are few-shot learners","author":"Brown","year":"2020","journal-title":"ArXiv, abs\/2005.14165."},{"key":"ref7","article-title":"Evaluating large language models trained on code","volume-title":"arXiv","author":"Chen","year":"2021"},{"key":"ref8","first-page":"4299","article-title":"Deep reinforcement learning from human preferences","volume":"30","author":"Christiano","year":"2017","journal-title":"Adv. Neural Inf. Proces. Syst."},{"key":"ref9","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/2024.acl-long.70","article-title":"Deepseekmoe: towards ultimate expert specialization in mixture-of-experts language models","volume-title":"arXiv","author":"Dai","year":"2024"},{"key":"ref10","article-title":"The llama 3 herd of models","volume-title":"arXiv","author":"Dubey","year":"2024"},{"key":"ref11","doi-asserted-by":"publisher","first-page":"52","DOI":"10.1038\/s41537-023-00379-4","article-title":"ChatGPT: these are not hallucinations \u2013 they\u2019re fabrications and falsifications","volume":"9","author":"Emsley","year":"2023","journal-title":"Schizophr"},{"key":"ref12","article-title":"An introduction to reinforcement learning: fundamental concepts and practical applications","volume-title":"arXiv","author":"Ghasemi","year":"2024"},{"key":"ref13","article-title":"Better & faster large language models via multi-token prediction","volume-title":"arXiv","author":"Gloeckle","year":"2024"},{"key":"ref14","year":"2024"},{"key":"ref15","article-title":"Deepseek-r1: incentivizing reasoning capability in llms via reinforcement learning","volume-title":"arXiv","author":"Guo","year":"2025"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2401.14196","article-title":"DeepSeek-coder: when the large language model meets programming - the rise of code intelligence","author":"Guo","year":"2024","journal-title":"arXiv"},{"key":"ref17","first-page":"238","article-title":"ChatGPT and the Chinese room argument: an eloquent AI conversationalist lacking true understanding and consciousness","author":"Hamid","year":"2023"},{"key":"ref18","first-page":"85","article-title":"Beyond probabilities: unveiling the delicate dance of large language models (LLMs) and AI-hallucination","author":"Hamid","year":"2024"},{"key":"ref19","first-page":"506","article-title":"Reinforcement learning and attractor neural network models of decision-making","volume-title":"Computational Models of the Brain","author":"Hamid","year":"2019"},{"key":"ref20","article-title":"Unlocking the potential of ChatGPT: a comprehensive exploration of its applications, advantages, limitations, and future directions in natural language processing","volume-title":"arXi","author":"Hariri","year":"2023"},{"key":"ref21","article-title":"Measuring massive multitask language understanding","volume-title":"arXiv","author":"Hendrycks","year":"2020"},{"key":"ref22","doi-asserted-by":"publisher","first-page":"100583","DOI":"10.1016\/j.taml.2025.100583","article-title":"Deepseek vs. chatgpt: a comparative study for scientific computing and scientific machine learning tasks","volume":"15","author":"Jiang","year":"2025","journal-title":"arXiv"},{"key":"ref23","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/D18-2012","article-title":"Sentencepiece: a simple and language independent subword tokenizer and detokenizer for neural text processing","author":"Kudo","year":"2018"},{"key":"ref24","article-title":"Training language models to self-correct via reinforcement learning","volume-title":"arXiv","author":"Kumar","year":"2024"},{"key":"ref25","article-title":"Gshard: scaling giant models with conditional computation and automatic sharding","volume-title":"arXiv","author":"Lepikhin","year":"2020"},{"key":"ref26","article-title":"Deepseek-v3 technical report","volume-title":"arXiv","author":"Liu","year":"2024"},{"key":"ref27","first-page":"864","article-title":"DeepSeek vs. ChatGPT: comparative efficacy in reasoning for adults\u2019 second language acquisition analysis","volume-title":"Humanities and Educational Studies Journal","author":"Mahyoob","year":"2025"},{"key":"ref28","doi-asserted-by":"publisher","first-page":"1458707","DOI":"10.3389\/frai.2025.1458707","article-title":"Detection and classification of ChatGPT-generated content using deep transformer models","volume":"8","author":"Maktabdar Oghaz","year":"2025","journal-title":"Front. Artif. Intell."},{"key":"ref29","article-title":"ChatGPT vs. DeepSeek: a comparative study on AI-based code generation","volume-title":"arXiv","author":"Manik","year":"2025"},{"key":"ref30","doi-asserted-by":"crossref","DOI":"10.1109\/FCCM48280.2020.00012","article-title":"Accelerating proximal policy optimization on CPU-FPGA heterogeneous platforms","author":"Meng","year":"2020"},{"key":"ref31","doi-asserted-by":"crossref","DOI":"10.1101\/2025.01.27.25321169","article-title":"Comparative evaluation of advanced AI reasoning models in pediatric clinical decision support: ChatGPT O1 vs. DeepSeek-R1","volume-title":"medRxiv","author":"Mondillo","year":"2025"},{"key":"ref32","doi-asserted-by":"crossref","DOI":"10.36227\/techrxiv.173273427.76836200\/v1","article-title":"ChatGPT is all you need: untangling its underlying AI models, architecture, training procedure, capabilities, limitations and applications","volume-title":"TechRxiv.","author":"Naik","year":"2024"},{"key":"ref33","doi-asserted-by":"publisher","first-page":"210","DOI":"10.7326\/M23-2772","article-title":"Large language models in medicine: the potentials and pitfalls","volume":"177","author":"Omiye","year":"2023","journal-title":"Ann. Intern. Med."},{"key":"ref34","year":"2022"},{"key":"ref35","year":"2023"},{"key":"ref36","year":"2024"},{"key":"ref37","first-page":"27730","article-title":"Training language models to follow instructions with human feedback","volume-title":"Adv. Neural Inform. Process. Syst.","author":"Ouyang","year":"2022"},{"key":"ref38","doi-asserted-by":"publisher","first-page":"9780","DOI":"10.1609\/aaai.v33i01.33019780","article-title":"Meaningful explanations of black box AI decision systems","volume":"33","author":"Pedreschi","year":"2019","journal-title":"Proc. AAAI Conf. Artif. Intell."},{"key":"ref39","year":"2024"},{"key":"ref40","doi-asserted-by":"crossref","DOI":"10.18653\/v1\/P16-1162","article-title":"Neural machine translation of rare words with subword units","author":"Rico Sennrich","year":"2016"},{"key":"ref41","doi-asserted-by":"crossref","DOI":"10.20944\/preprints202502.1608.v1","article-title":"Comprehensive analysis of transparency and accessibility of chatgpt, deepseek, and other sota large language models","volume-title":"arXiv","author":"Sapkota","year":"2025"},{"key":"ref42","article-title":"Proximal policy optimization algorithms","volume-title":"arXiv","author":"Schulman","year":"2017"},{"key":"ref43","article-title":"Deepseek-v2: a strong, economical, and efficient mixture-of-experts language model","volume-title":"arXiv","author":"Shao","year":"2024"},{"key":"ref44","article-title":"Deepseekmath: pushing the limits of mathematical reasoning in open language models","volume-title":"arXiv","author":"Shao","year":"2024"},{"key":"ref45","article-title":"Fast transformer decoding: one write-head is all you need","volume-title":"arXiv","author":"Shazeer","year":"2019"},{"key":"ref46","article-title":"Learning to summarize from human feedback","volume-title":"arXiv","author":"Stiennon","year":"2020"},{"key":"ref47","doi-asserted-by":"publisher","first-page":"127063","DOI":"10.1016\/j.neucom.2023.127063","article-title":"Roformer: enhanced transformer with rotary position embedding","volume":"568","author":"Su","year":"2024","journal-title":"Neurocomputing"},{"key":"ref48","article-title":"Towards smaller, faster decoder-only transformers: architectural variants and their implications","volume-title":"arXiv","author":"Suresh","year":"2024"},{"key":"ref49","volume-title":"Reinforcement learning: An introduction","author":"Sutton","year":"2018"},{"key":"ref50","first-page":"5998","article-title":"Attention is all you need","volume":"30","author":"Vaswani","year":"2017","journal-title":"Adv. Neural Inform. Process. Syst."},{"key":"ref51","article-title":"Auxiliary-loss-free load balancing strategy for mixture-of-experts","volume-title":"arXiv","author":"Wang","year":"2024"},{"key":"ref52","article-title":"Math-shepherd: a label-free step-by-step verifier for llms in mathematical reasoning","volume-title":"arXiv","author":"Wang","year":"2023"},{"key":"ref53","first-page":"39957","article-title":"Ar-diffusion: auto-regressive diffusion model for text generation","volume":"36","author":"Wu","year":"2023","journal-title":"Adv. Neural Inform. Process. Syst."},{"key":"ref54","article-title":"Introduction to transformers: an NLP perspective","volume-title":"arXiv","author":"Xiao","year":"2023"},{"key":"ref55","article-title":"DeepSeek-coder-V2: breaking the barrier of closed-source models in code intelligence","volume-title":"arXiv","author":"Zhu","year":"2024"}],"container-title":["Frontiers in Artificial Intelligence"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2025.1576992\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,6,19]],"date-time":"2025-06-19T05:30:13Z","timestamp":1750311013000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2025.1576992\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,6,19]]},"references-count":55,"alternative-id":["10.3389\/frai.2025.1576992"],"URL":"https:\/\/doi.org\/10.3389\/frai.2025.1576992","relation":{},"ISSN":["2624-8212"],"issn-type":[{"value":"2624-8212","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,6,19]]},"article-number":"1576992"}}